Method and a system for automatic labeling of activity on ECG data

1. A method for automatic labeling of a plurality of activities of a subject using electrocardiography (ECG) data, the method comprising: sensing, with a wearable medical device, an ECG signal having a plurality of segments, the plurality of segments including both a ST segment and a PQ segment; conditioning, with an adaptive band pass filter, the sensed ECG signal; subsequent to the conditioning with the adaptive band pass filter, identifying, with a processor, a plurality of fiducial points in the plurality of segments of the sensed ECG signal; performing, with the processor, after identifying the plurality of fiducial points, fragment analysis on the segments of the sensed ECG signal, wherein the fragment analysis includes extraction of feature vectors from both the ST segment and the PQ segment; generating a trained classifier model based on the fragment analysis of the sensed ECG signal; mapping, with the trained classifier model, the plurality of activities based upon the sensed ECG signal; obtaining, with the processor, a respiration rate based upon both the plurality of fiducial points and the fragment analysis; differentiating, with the processor, based upon the respiration rate, the plurality of activities performed by the subject; identifying, with the processor, a particular activity of the plurality of activities based upon variation in the plurality of fiducial points in the sensed ECG signal, the differentiating, and morphological changes in the ECG signal; and storing, with the processor after identifying the particular activity, the sensed ECG signal with a label corresponding to the identified activity.

2. The method as claimed in claim 1 , further comprising: fitting the ST segment and the PQ segment of the sensed ECG signal based on cubic spline fitting to provide a fitted segment of the ECG signal.

3. The method as claimed in claim 2 , further comprising: applying a Karhunen Loeve Transform (KLT) on the fitted segment of the sensed ECG signal to obtain principal component values.

4. The method as claimed in claim 3 , further comprising: computing eigen values of the principal component values to obtain a feature set.

5. The method as claimed in claim 4 , further comprising: classifying the feature set based on the trained classifier model to map the feature set to a corresponding activity of the plurality of activities.

6. The method as claimed in claim 1 , wherein the trained classifier model provides a basis for mapping different activities of the plurality of activities in correspondence with the sensed ECG signal.

7. The method as claimed in claim 6 , further comprising: marking the identified activity on the sensed ECG signal automatically, wherein the marking step corresponds to the stored label.

8. A system for automatic labeling of a plurality of activities of a subject using ECG data, the system comprising: a wearable medical device configured to sense an ECG signal having a plurality of segments, the plurality of segments including both a ST segment and a PQ segment; an adaptive band pass filter configured to condition the sensed ECG signal; a processor configured to identify, subsequent to the conditioning with the adaptive band pass filter, a plurality of fiducial points in the plurality of segments of the sensed ECG signal; perform fragment analysis, after identifying the plurality of fiducial points, on the segments of the sensed ECG signal, wherein the fragment analysis includes extraction of feature vectors from both the ST segment and the PQ segment; generate a trained classifier model based on the fragment analysis of the sensed ECG signal; map, with the trained classifier model, the plurality of activities based upon the sensed ECG signal; obtain a respiration rate based upon both the fiducial points and the fragment analysis; differentiate based upon the respiration rate, the plurality of activities performed by the subject; identify a particular activity of the plurality of activities based upon variation in the plurality of fiducial points in the sensed ECG signal, the differentiating, and morphological changes in the sensed ECG signal; and store, after identifying the particular activity, the sensed ECG signal with a label corresponding to the identified activity.

9. The system as claimed in claim 8 , wherein the processor is further configured to fit the ST segment and the PQ segment of the sensed ECG signal based on cubic spline fitting to provide a fitted segment of the sensed ECG signal.